Waiting until December could cost you more. Get ahead of year-end today. →

What a 95-year old law firm’s data is worth

What is a 95-Year-Old Law Firm's Data Worth?

A 95-year-old law firm is winding down, and an AI company reportedly wants its data. Here is what a firm's archive holds, what it could potentially be worth, and why legal data needs a higher privacy bar than almost anything else.

Weil, Gotshal & Manges recently told its firm partners it plans an orderly wind-down. A day later, The Information reported that legal AI startup Crosby had submitted a letter of intent for some of the firm's assets, including its internal data.

When Dewey & LeBoeuf, a firm of similar size, collapsed in 2012, the main story was about partners, clients, and creditors. Data wasn't part of it. Today it's a bid line item: Spirit Airlines' internal emails and chats drew offers from Google, Mercor, and micro1 in bankruptcy court.

Through SimpleClosure’s AssetHub, 350+ companies have brought their code and workspace data to AI buyers. Here's how we'd think about a law firm’s data.

How much data does a law firm create?

A firm of Weil's size and age likely holds hundreds of millions of emails and tens of millions of documents. Weil was founded in 1931 and has about 1,200 attorneys. Our back-of-envelope estimate of its archive:  

Data type

Estimate

How we got there

Emails

500M to 1B

~150 a day per lawyer, ~60 per staffer, if ~20 years retained

Documents and versions

75M to 100

Large firms typically hold tens of thousands of documents per lawyer in their document management systems

Agreement drafts and redlines

15M+

Assuming ⅕  of documents are contract drafts

Time entries

50M+

Each one a short description of a task plus its duration

To be clear, these are our (very) rough estimates, not figures from Weil.

What could it be worth?

The benchmark: Google won Spirit's auction at $10 million for what was described at the time as about 100 million emails, 500 million Teams items, about 38 million OneDrive and SharePoint files and 30 million lines of code. Mercor bid $7.5 million, and micro1 later offered $12.5 million. 

That works out to roughly $600 to $700 per employee, or about two cents per message in the original bid. Applied to a law firm, the quality of data drives the value much higher. Our buyers tell us legal and professional services are some of the most difficult yet valued verticals to secure data from because of the context their authors provide. For example, a partner's markup on a merger agreement carries greater expert signal than a gate-change chat. 

Revised sale terms filed on November 9 show a much larger archive: nearly 2 billion emails, 664 million Teams items, and 91 million files. The price stayed at $10 million, and in it Google covers the cost of 3rd party de-identification. The sale still awaits court approval, but this shows that pricing per email or document doesn't hold up. Per message, Spirit's price fell from about two cents to under half a cent as the count grew, and the $10 million didn’t move. Buyers price a company's full record of how work gets done, which is why archives of similar size can be worth very different amounts. Not every bankruptcy archive is a Spirit. Weil's matches Spirit's on documents (75M to 100M versus 91M files). In place of Spirit's 30 million lines of code, it holds 15M+ drafts and redlines and 50M+ time entries, the legal equivalent of a codebase and its commit history. The quality of that data is also higher. 

But most of a law firm's archive is client work… 

Once you account for what can responsibly be licensed, including matters that can't be de-identified or cleared, discovery materials under protective orders, third-party material and email deleted under retention policies, we'd expect only a small part of the entire corpus to make it through, bringing the value of a one-time, upfront exclusive license closer to $1M to $2M. This is a very rough estimate, but generally aligns with what we’re seeing in the market.

Why legal data needs a higher bar for de-identification

Masking names at a corpus of this size is not easy, but it is doable. In legal data however, especially of this size, the deal itself is often the identifier. It’s difficult to “find-and-replace” the path to privacy when the deal made the front page.

De-identification technology replaces every person and company name with a token. The tooling here is quickly maturing: Spirit hired Tonic.ai to scrub its data, and the court's privacy ombudsman recommended approval. At SimpleClosure, we’ve been working on de-identification for months, specializing in startup data. 

But a law firm’s data is different, because its most important work is often public. Mask "Microsoft" and "Activision," and a $69 billion video game acquisition challenged by the FTC is still recognizable. A landmark bankruptcy stays famous even with every name removed.

At Spirit, the flight attendants' union argued that individuals could be re-identified from the links between datasets. Legal data raises that risk further. We propose a higher standard that looks like this:

  • Mask every entity and person, including in signature blocks, metadata, file names, and email headers.

  • Band the numbers, so deal values, fees and settlements become ranges.

  • Shift dates, so timelines keep their shape but lose their calendar.

  • Carve out marquee matters entirely: any deal or case covered by the press.

  • Remove docket and filing references that point to a single case.

  • Bar re-identification by contract, and audit buyers against it.

What's in, what's out

The most licensable data is what the firm owns outright. Client work needs a stricter path, and some of what sits on a firm's servers is never the firm's to sell.

Firm-owned. The precedent and templates bank, knowledge-management memos on points of law, training materials, pitch decks and operational data such as finance reports and IT tickets.

Client work. A lawyer's duty of confidentiality is broader than privilege: it covers everything about a representation. Sharing client files with a non-lawyer buyer requires informed client consent. Timing matters too. In a wind-down, files follow clients to their new counsel, so the clearable pool shrinks every week.

3rd party work. Licensed form libraries built into precedent banks, opposing counsel's drafts inside redline chains, and vendor work products are likely to be excluded in sales. Spirit faced the same issue when software vendor Springshot objected that the archive held data its platform generated.

Always excluded. Just as with AssetHub, excluding sensitive employee, compensation and health systems information is standard practice.

Who's buying, and why the drafts matter most

The value isn't the signed agreement, it's the path to it.

Final agreements for public deals often end up in SEC filings. What nobody has is the sequence: draft, markup, counter, final, with the emails explaining each move. That record of how expert work actually gets done, including what a senior partner changes, and what gets conceded and why is the most valuable data.  Time entries add a duration to every step, which is exactly what buyers are looking for.

There are multiple categories buyers fall into, each of which are looking for this type of data:

  • Legal AI companies. AI-native players like Crosby, Harvey, Legora and Spellbook need examples of real negotiation to train and evaluate their agents.

  • Legal information incumbents. Clio paid $1B for vLex, whose platform runs on more than a billion legal documents of largely public law. Private deal process data is missing from its stack.

  • Frontier AI labs. Google has already demonstrated it will bid for enterprise data as it did with Spirit Airlines.

  • RL environment companies. Data brokers like Mercor and micro1 have vast orders to fill to support reinforcement learning environments’ efforts to make agents truly autonomous for long durations.

With many different types of buyers, deal structure matters significantly. One exclusive sale is an option, but non-exclusive licenses to several buyers often return more. With SimpleClosure’s AssetHub, we give operational sellers not only the best returns on multiple non-exclusive licenses to each of the groups above, but also ensure they do so responsibly.  And with company dissolutions, we help maximize an initial exclusive deal for their assets that returns the most to stakeholders and allows a compliant shut down. 

Your archive is a record of how expert work gets done

Any firm that has operated for decades holds assets AI buyers can't get anywhere else. The question is no longer whether they’re worth anything, it's which assets can be licensed responsibly, and quickly enough, before the files walk out the door.

This challenge applies well beyond law firms. Accounting, consulting, engineering, and financial services firms all hold the same kind of record.

So, what is your firm's data worth? Get a complimentary asset valuation →

Preview of a completed SimpleClosure dissolution

Learn how SimpleClosure can help you shutdown your startup hassle-free

Schedule a call

It's time to get the closure you deserve.

Tell us about your business, we'll build your custom dissolution plan.